Lexical Knowledge of monolingual and bilingual children

  • Upload
    tom-way

  • View
    225

  • Download
    0

Embed Size (px)

Citation preview

  • 8/22/2019 Lexical Knowledge of monolingual and bilingual children

    1/20

    Lexical Knowledge of Monolingual and BilingualChildren

    MARIANNE VERHA LLEN and ROB SCHOONENUniversity of Amsterd am 1

    The aim of the present study is to gain insight into the lexico-sem anticknowledge of bilingual children growing up in a second-language immersionenvironment. The research focus is on aspects of lexical knowledge that arerelevant for school success. Data were obtained by asking 40 monolingualDu tch and 40 bilingual Turkish children (9 and 11 year olds) to explain themeanings of comm on Dutch nouns in an extended word definition task. In ahighly structured interview session the children were stimulated to express allthe meaning aspects they could think of.We evaluated both the differences between the two ethnic groups and theeffect of age in relation to the types of meaning the children expressed, by meansof statistical (loglinear) model fitting. Important differences were found withrespect to the number of meaning aspects expressed and with respect to the

    nature of meaning relations involved. Com pared to the monolingual Dutchchildren, the bilingual Turkish children tended to allot less extensive and lessvaried meanings to Dutch words.I N T R O D U C T I O NIn sociolinguistics, the relationship between children's language capacity andtheir educational achievement has been a frequent subject of discussion andinvestigation since the 1960s. Although the exact relationship is still not fullyunderstood, it is widely acknowledged that certain aspects of languageproficiency are of great relevance for academic achievement. Lexical know-ledge is assumed to be one of the most important determinants. As Dickinson(1984:359) s ta tes :

    Educators have long known that the size of children's vocabularies correlates withgeneral intelligence scores, reading ability and school success.Several investigations have made clear that word knowledge is strongly relatedto reading comprehension (for a survey, see Beck, McKeown, and Omanson1987), clarifying the relationship between a child's lexical knowledge andschool performance.Next to studies with monolingual children as subjects, research on thelanguage proficiency of bilingual children has become increasingly important.In Dutch primary education, around 10 per cent of the school populationconsists of children from ethnic minority groups (cf. Boogaard, Damhuis, DeGlopper, and Van den Berg 1990). Many of these children have a Turkish orMoroccan background. As a rule, minority children attend Dutch schools inApplied Linguistics, Vol. 14. No . 4 Oxford University Press 199 3

    atOUPsiteaccessonJuly31,2012

    http://applij.oxfordjournals.org/

    Downloadedfrom

    http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/
  • 8/22/2019 Lexical Knowledge of monolingual and bilingual children

    2/20

    MARIANNE VERHALLEN AND ROB SCHOONEN 345mixed classes, together with monolingual Dutch peers, and take part in theregular Dutch curriculum. In primary education, however, minority children arerelatively less successful. We see that disproportionate numbers attend lowerforms of secondary education. Also they leave secondary education withoutqualifying in far greater numbers than their Dutch peers.One of the main reasons for the low educational achievement of the minoritychildren is the fact that many do not have a sufficient command of the Dutchlanguage. Most of the non-native children have been born and brought up inThe Netherlands, but the language of their primary socialization is the minoritylanguage. Linguistic research has shown that they are far less proficient in Dutchthan their native Dutch classmates, with their most prominent delay at thelexical level. The Dutch vocabulary of four-year old immigrant children is verylimited at the start of their school career. This would not be so serious if theirlexical delay diminished during the elementary school period, but researchshows that lexical differences between Dutch and immigrant children evenincrease as the children grow older (Verhoeven and Vermeer 1989).Up to now, most research has concentrated on the macro-level of the lexicon(knowledge in breadth), focusing on questions like 'how many words have beenacquired?'. Studies on the micro-level (knowledge of individual words in depth),concentrating on questions like 'how well do children know the Dutch words?'have hardly been taken up as yet.In this article, we present the results of a research project on these 'micro-level' aspects of the lexical knowledge of non-native children learning the Dutchlanguage. The study deals with Turkish children who had participated in theDutch educational system from the age of four. Before discussing the contentand results of this research, we would like to discuss some general features oflexical development in relation to academic achievement, and the connectionbetween lexical and conceptual development.The acquisition of the (L2) lexiconThe growth of the lexicon not only concerns the acquisition of m ore and morewords, but also the acquisition of multiple meanings assigned to words. Wordsrarely have one fixed meaning. Even words with an apparently unambiguousmeaning comprise different aspects of meaning. For example, the various facetsof the meaning potential of words like 'piano ' can be highlighted by putting theword in different sentences (Barclay, Bransford, Franks, McCarrell, and Nitsch1974:472):

    1. the man lifted the piano2. the man tuned the piano3. the man smashed the piano4. the man sat on the piano5. the man photographed the piano

    In these different contexts, various meaning aspects of the word 'piano' areemphasized: 'is heavy', 'musical', 'made largely of wood', 'has a flat surfaceperpendicular to gravitational pull', and 'has a characteristic shape'.

    atOUPsiteaccessonJuly31,2012

    http://applij.oxfordjournals.org/

    Downloadedfrom

    http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/
  • 8/22/2019 Lexical Knowledge of monolingual and bilingual children

    3/20

    346 LEXICAL KNOWLEDGE IN CHILDRENIf people are asked to express their full knowledge of a simple word they cangive an impressive amount of meaning aspects, as Anglin (1985) showed. Hedrew up a schem e representing the aspects of mean ing expressed by one ad ult in

    a definition of the wo rd 'do g' (see Figure 1).The different aspects of meaning of individual words are acquired step bystep as children encounter words in different contexts and situations. Sternbergpoints to context as ' the major so urce of grow th in vocabulary know ledge' (seeSternberg 1987). Children need to experience words in different contexts toacquire all possible facets of the potential meaning. Nagy, Herman, andAnderson (1985) show that any one encounter with a word usually results inonly a small gain in knowled ge. Th is proc ess of voc abu lary acqu isition pro ce ed sin small steps, each step contributing to the development of an extensivemeaning potential .It is well known that monolingual children are continuously confronted withwo rds in a great variety of contex ts and situation s in and ou t of schoo l, whe reasbilingual children in an L2-submersion environment (like the linguistic minoritychildren in the Dutch educational system), have more restricted languageexp eriences in the L2 : they dep en d a lmo st exclusively on the L2 inp ut theyreceive at school, and therefore en cou nter wo rds in less diverse situations thantheir monolingual peers. As a result of this, it is to be expected that bilingual

    children will eventually allot a less exten sive an d less varied ran ge of mean ing to(L2) words than monolingual children.It is imp ortan t to take a closer look at the se differences be cau se kno wle dg e ofword m eanings plays an im portan t role not only in everyday com m unication butmost significantly in the acquisition of academic knowledge in education.

    Relation of word knowledge and academic achievementA t school, children are not only con fronted with new wo rds and new co nce pts,but also with new meaning relations between words and concepts. Youngchildren are familiar with w ords like 'ros e', 'tulip', 'flower', 'plant'; this doe s notmean, howev er, that they are aware of the hierarchical relations between theseconcepts; it is only later on that they learn the relationship between the words' rose ' and ' tulip ' as co-ordinated co ncep ts , both dom inated and supero rdinatedby the conc ept 'flower', w ith the con ce pt 'plan t' still higher in the hiera rchy . W efind an extensive description of the development of meaning structures inVygotsky (1 96 2). Vygotsky points ou t that the acquisit ion of word me anings ismore than a simple clustering of mean ing asp ects. H e conclu des that:

    ... word meanings evolve. When a new word has been learned by the child, itsdevelopment is barely starting: the word at first is a generalization of the most primitivetype; as the child's intellect develops, it is replaced by generalizations of a higher andhigher typea process that leads in the end to the formation of true concepts.(Vygotsky 1962:83)The ability to handle abstract concepts is essential particularly in the upperlevels of primary education: whereas in nursery schools reference is made to a

    atOUPsiteacces

    sonJuly31,2012

    http://applij.oxfordjournals.org/

    Downloadedfrom

    http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/
  • 8/22/2019 Lexical Knowledge of monolingual and bilingual children

    4/20

    MARIANNE VERHALLEN AND ROB SCHOONEN 347

    animaldomesticanimal

    mammal pet

    DO G/-working dogs/ -guard < ogs/-hunting dogsL wild dogs

    mongrel/cross-breed

    - afghans- beagles-collies- boxers

    -Russian wolfhounds-dalmatians-bulldogs- great danes- dachshunds-German shepherds (Rin Tin Tin)- terriers

    carnivorous, but also eats grass.caninedomesticated from animals that formerly lived in the wild.wild dogs live in packs.requires care of humanspeople play with it: games of tossing an objectand having a dog retrieve itused for protection: chases away intruders in a dwelling

    /have to be walked^affectionate: lick one's face^given affection: scratching, petting,tak e care of biological needs,put out territorial markers-eating-excreting-sex~ roaming aroundsplaying with other dogs- exploring^sniffingMess intelligent than a human but more so than a fish^ you can train them^ they barkv has teeth

    has paws: claw, pad, nails^ has hair of different sorts

    facial structure such that face comes to a pointwith a nose at the end, but not always^sensitive nosetongue that hangs out when it needs more airsize ranging from 4' at shoulder to about 6" atshoulder, most dogs about 1 '6 " to 2'6" at shoulder

    , clipped ears or tail if domesticatedhas four legshas tail

    t has ey es. has heady has ear flapsi has mouthlhas genital organsihas less hair on underbelly

    there are different breeds for different purposesFigure I: Mea ning aspects of 'dog' (Anglin 1985)

    atOUPsiteacces

    sonJuly31,2012

    http://applij.oxfo

    rdjournals.org/

    Do

    wnloadedfrom

    http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/
  • 8/22/2019 Lexical Knowledge of monolingual and bilingual children

    5/20

    348 LEXICAL KNOWLEDGE IN CHILDRENparticular dog (one of the children's) as a pet, in the sixth grade lessons are aboutdogs as mammals. During successive school years, an extensive system ofeducational knowledge is built up that is represented in word meanings andmeaning relations. Attribution of meaning grows more complex, whileprinciples like 'categorization' and 'abstraction' are further extended. Thus,lexical development comprises more than extending vocabulary size and (at thelevel of individual words) extending knowledge of meaning components. It isalso a gradual structuring process, in which higher-level interrelationshipsbetween words and their meanings are established.To find an answer to the question 'how well do children know the w ords theyknow', both quantitative dimensions (with respect to the amount of meaningaspects acquired) and qualitative dimensions (with respect to the nature of themeaning relations involved) must be assessed. It is important to investigate bothdimensions of the meaning system to find out if there are differences betweenbilingual and monolingual children.Possible differences in application of meaning can be assessed by settingchildren 'extended' definition tasks. In such tasks, children are asked to expressall the meaning aspects of a word they can think of. In the next section, we willdiscuss the advantages of this method.Research in word knowledgeTo m easure vocabulary size, children are usually set formal lexical tests (p icturetests, multiple-choice tests, true-false type tests). In such tests, ch ildren can onlydemonstrate that they either know a word or that they do not. The fact thatwords can be known to a smaller or larger extent is ignored. Cronbach (1942)has pointed out that there are various degrees in the knowledge of the meaningsof words, but that these differences are not assessed in formal vocabulary tests.Definition tasks, as opposed to formal vocabulary tests, give the opportunity tostudy lexical know ledge in rather more depth . They are a means to establish howwell or how extensively a word has been acquired. A particular advantage of thedefinition method is that we can study lexical development in the light ofcognitive development. The analysis of word definitions can give us someunderstanding of the way in which word meanings and meaning structures arementally represented.Some studies (cf. Anglin 1985) have shown important qualitative changes inthe development of word meanings in LI by applying this method. Oneimportant aspect concerns the development of paradigmatic sense relations.Both in semantics and theories of lexical development (cf. Vygotsky 1962;Miller and Johnson-Laird 1976; Kuczaj 1982; Cruse 1986) a fundamentaldistinction is made between paradigmatic sense relations and syntagmatic senserelations. Paradigmatic relations are represented as vertical relations becausethey refer to hierarchical relations, in contrast with horizontal, syntagmaticrelations. These hierarchies can be taxonomic when referring to class-inclusion,and partonom ic when referring to part-whole relations (cf. Cruse 1986).The importance of the paradigmatic relations lies in the fact that it is here that

    atOUPsiteaccessonJuly31,2012

    http://applij.oxfordjournals.org/

    Downloadedfrom

    http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/
  • 8/22/2019 Lexical Knowledge of monolingual and bilingual children

    6/20

    MARIANNE VERHALLEN AND ROB SCHOONEN 349

    Taxonomic hierarchy (Cruse 198 6:136)creature

    animal birddog elephant robin ea

    spaniel alsatian

    fish insectcod trout ant butterfly

    Partonomic hierarchy (Cruse 198 6:15 7)body

    head neck trunk arm legforearm hand

    palm fingerFigure 2: Paradigmatic relations

    'categorization', 'abstraction', and 'generalization' (see above-mentioned quota-tion of Vygotsky 1962) are realized. In word definition tasks, subjects aredirectly asked to produce paradigmatic relations, especially the taxonomictypes expressing abstract categorization: a formal definition practically alwaysfirst indicates to which category the definiendum belongs ('a spaniel is a dog that. . . , a robin is a bird that ..., a rose is a flower that ...', etc.). In contrast toparadigmatic relations, syntagmatic relations are not hierarchicali.e. they arehorizontal; for example, 'a rose grows on a bush', or 'a rose is red'.Drawing on this distinction, we set up a classification scheme, in which dif-ferent types of meaning aspects were distinguished. A description of this modelis given in the next section. First, we give an outline of our research design.METHOD

    Subjects and proceduresThe subjects were 80 children at the ages of 9 and 11comprising 40Dutchchildren and 40 Turkish children, equally distributed between the two age-groups (altogether: 2X2 = 4 groups). All children came from (three) schoolswith high percentages of minority children (over 50per cent). All the schoolswere located in the same lower-class neighbourhood. The participating childrenwere selected with the help of their teachers, the criteria for selection being thatthey should be born in The Netherlands, and that their language proficiency andtheir academic results should not be deviantly high or low.By means of a receptive-vocabulary picture test, we divided each group intohigh lexical achievers and low lexical achievers. Children scoring above the

    atOUPsiteaccessonJuly31,2012

    http://applij.oxfordjournals.org/

    Downloadedfrom

    http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/
  • 8/22/2019 Lexical Knowledge of monolingual and bilingual children

    7/20

    350 LEXICAL KNOWLEDGE IN CHILDRENmedian of their ethnic group were considered high proficient, the childrenscoring below the median were considered low proficient.1Each child took part in an individual, highly structured interview session thatwas recorded on tap e. The children were given six stimulus words and asked togive as many meaning aspects as they could think of. H ere we report on theseinterviews. The p resent study, however, is part of a larger project in which theTurkish children were also interviewed in Turkish for the same w ords.

    It must be stressed that we did not just elicit formal definitions (with questionslike 'what does the word ... mean?', or 'what is a . . .? ') . In order to cover allpossible meaning dimensions, the subjects were confronted with variousstimulus questions:1. (wat betekent. .. I watiseen...)[what does . . . mean / what is a . . .]2. (hoe zou je uitleggen wat een...is)[how wou ld you expl ain wha t a . . . is]3 . (wat zie/e aan een . ..)[what do you see if you look at a .. .]4. (wat voorsoorten .. .zijn er)[what kin ds o f . .. are there]5 . (wat voor soort ding/lets is een . ..)[what kind of thing is a .. .]6. (wat kun/moetje doen met een . ..)[what can/m ust you do with a .. .]7. (kunje 3 zinnetjes met het woord . .. maken)[can you make 3 sentences with the word . . . ]Each answer to the first 6 questions was followed by extra elicitation ('can youtell me mo re?' or 'yes, and . . .? '), so for each word at least 13 standard questionswere asked.

    In total, 6 different stimulus words were presented: neus (nose), roofdier(predator/beast of prey), wekker (alarm clock), geheim (secret), boek (book),and haar (hair).We expected all children to be familiar with these stimulus words. However, itturned out that in the Turkish group 6 nine-year old children did not know theword roofdier. To 3 other Turkish children the word wekker was unknown (2nine-year olds and 1 eleven-year old). These missing data have been m ade upfor by converting the available data to a standard group size of 10.2ScoringTo analyse the word meanings expressed in the definition task we developed aclassification model, based on current semantic theories (cf. Miller andJohnson-Laird 1976; Cruse 1986) and on results of investigations into thelexical development of children (Vygotsky 1962; Kuczaj 1982; Anglin 1985).The classification model contains six different categories of meaning relations.We took meaning relations as a starting point because all different types of

    atOUPsiteaccessonJuly31,2012

    http://applij.oxfo

    rdjournals.org/

    Do

    wnloadedfrom

    http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/
  • 8/22/2019 Lexical Knowledge of monolingual and bilingual children

    8/20

    MARIANNE VERHA LLEN AND ROB SCHOONEN 351meaning aspects that children assign to words in definition tasks are expressedas meaning relations (for example, in a definition of 'spaniel', the meaning aspect'dog' is made explicit by expressing the meaning relation 'a spaniel is adog'). Assuch, meaning aspects can be identified according to the meaning relationsinvolved.The main categories and the subsequent divisions are represen ted in Figure 3.It should be stressed that the scheme presented in Figure 3 was developed toidentify and count the different sorts of meaning aspects that children tend touse in the definition task. It was developed to be used as a working model, not asa description of a semantic theory.In the model, the first basic categories of paradigmatic and syntagmaticrelations are broken down into a number of subcategories.In the paradigmatic category three different subcategories were distin-guished. In addition to the two types (taxonomical and partonomical) that wediscussed earlier, a separate category was postulated for 'dummy super-ordinates'. Dummy superordinates are frequently used by children in defini-tions. Watson (1985) pointed to the fact that children tend to substitute adummy when they are unable to find a proper superordinate term in formaldefinitions. Then they use 'empty' words like 'something' or 'thing' in order to fillthe formal definition-scheme. For example, they might say 'a cat is somethingthat has fur' instead of'a cat is an animal that has fur' (Watson 19 85 :18 9) .

    In the syntagmatic category a distinction is made between on the one hand'specific episodic' meaning relationsspatial and perceptualand 'association'on the other hand. The subcategory association is an umbrella-term for adiversity of meaning relations like associative, functional, and instrumentalsense relations. In this category it is difficult to establish appropriate dem arca-tions ('a nose is to smell, to breathe, to sniff, to pick', or 'you can catch a cold withyour nose').In our classification model, a third main category has been adopted; in

    analysing definitions it is important to distinguish between objective and

    I Paradigmatic

    II Syntagmatic

    III S ubjective -

    1.0 Hierarchical Dummy-

    1.1 Hierarchical Taxonomical

    1.2 Hierarchical Partonom ical -

    2.1 Specific Episodic-

    2.2 Association -3. Attitudes

    - empty superordinatesuperordinatesynonymsubordinate

    -constituents- spatial (location)-perceptual-associativespersonal opinions

    Figure 3 : Classification scheme of meaning aspects

    atOUPsiteaccessonJuly31,2012

    http://applij.oxfordjournals.org/

    Downloadedfrom

    http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/
  • 8/22/2019 Lexical Knowledge of monolingual and bilingual children

    9/20

    352 LEXICAL KNOWLEDGE IN CHILDRENsubjective m eaning relations. As opposed to the objective meaning aspects, sub-jective meaning aspects ('a nose is funny') are classified as 'attitudes'. Strictlyidiosyncratic meanings ('a nose is to put a clothes-peg on, like uncle Harryalways does for fun') were classified separately.All responses of the children were recorded on tape and scored according toour classification scheme. Inter-scorer reliability on a subset of the data wasestimated .90 (Cohen's kappa) and proved to be satisfactory.For each meaning aspect, expressed values were assigned with regard to thefollowing variables:E = Ethnic Background 1) Turkish2) DutchP = Lexical proficiency 1) Low2) HighA = Age 1) 9 years2) 11 yearsW = Word 1) Neu s(Nose)2) Roofdier (Predator)3) Wekker (Alarm clock)4) Geheim (Secret)

    5) Boek(Book)6) Haar(Hair)C = Category 1) Dummy superordinate2) Hierarchical taxonomical3) Hierarchical partonomical4) (Syntagmatic) specific5) Associations6) AttitudesThis design gave us the opportunity to check the effects of the subject variables(ethnic background, age, lexical proficiency) on the quality of the given defini-tions (i.e. the frequency with which certain meaning categories were used). Thedesign also yielded useful information about the influence of the stimulus word.STATISTICAL ANALYSISThe procedures described above led to 8,833 classified meaning aspects. Thefrequencies with which certain meaning aspects were mentioned can bedescribed in a five-way contingency table ( 2 X 2 X 2 X 6 X 6 ) : two ethnicbackgrounds (Dutch and Turkish); two proficiency levels (high = abovemedian, and low = below median of the ethnic group); two age groups (younger,9 years, and older, 11 years); six words, all of them nouns (nose, predator,alarm, secret, book, and hair); and six types of meaning relations (Dummy,Taxonomical, Partonomical, Syntagmatic specific episodic, Associative, andAttitudes).The question is which of these variables, especially the subject variables (age,

    atOUPsiteaccessonJuly31,2012

    http://applij.oxfordjournals.org/

    Do

    wnloadedfrom

    http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/
  • 8/22/2019 Lexical Knowledge of monolingual and bilingual children

    10/20

    MARIANNE VERHALLEN AND ROB SCHOONEN 353proficiency, and ethnic background), determine the number and the types ofmeaning aspects expressed by the children.For example, older children may be expected to mention more meaningrelations than younger children. In that case, we would find a so-called maineffect of Age (A). In addition, in the two age groups the expressed meaningaspects can be expected to be distributed in different proportions over thecategories. Older children are generally expected to use hierarchical para-digmatic meaning relations more frequently than younger children in theirdescriptions of word meaning. This association between Categories (C) andAge (A) leads to a so-called interaction effect between Categories andin thisexampleAge, notated as C*A.In a two, or three-way contingency table, it is relatively easy to test forassociations between categorical variables by means of the chi-square statistic.For complex contingency tables such as ours, loglinear model fitting, which canbe seen as a generalization of the chi-square test, is more suitable (Everitt 1976;Fienberg 1977). Loglinear model fitting for the analysis of categorical data iscomparable to an analysis of variance. Frequencies are described as theconsequence of main effects of variables and interactions. These models of mainand interaction effects can be evaluated statistically for their fit to the frequencydata. This evaluation leads to a 'scaled deviance', G2, which is approximatelychi-square distributed with df degrees of freedom.3 The difference between two(nested) models can also be evaluated statistically, since the difference in fit,(}diff(= Gf Gj), is also approximately chi-square distributed withdfdiff(= df, df2) degrees of freedom. If G ^ is statistically significant, it meansthat the (null) hypothesis that both models are equally good should be rejected.To find out which variables are relevant because of their main effects, i.e.causing differences in number of meaning aspects expressed, or because of theirinteraction effects, i.e. association with other variables, we fitted several modelsto the frequency data.

    Of special interest are, of course, those m odels in which the subject variablesAge, Proficiency, and Ethnic Background, or their interactions with Categories,are relevant param eters in the description of the data. Differences between thewords or between categories as such, or the interaction between Category andWord indicate respectively that (a) some words elicit more meaning aspectsthan others; (b) some categories are used more often in the meaning descrip-tions than others; or (c) different words elicit different kinds of descriptions.These effects are obvious and not at issue here . Our point of departure will be amodel containing these effects. This model which will be referred to as theminimal model postulates that all frequency data can be described in terms ofdifferences between W ords and between C ategories and an association betweenthese two variables, whereas the subject variables (Age, Proficiency, and EthnicBackground) would have no effect (see model 1 in Table 1). In other words, ifindeed the subject variables are of no influence on the kind of meaning aspectschildren express, the minimal model should fit the data. However, should thisminimal model (1) have to be rejected, then we can assume that (some of) the

    atOUPsiteaccessonJuly31,2012

    http://applij.oxfo

    rdjournals.org/

    Do

    wnloadedfrom

    http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/
  • 8/22/2019 Lexical Knowledge of monolingual and bilingual children

    11/20

    354 LEXICAL KNOWLEDGE IN CHILDRENsubject variables have an effect on the meaning aspects children express. In thatcase, we are able to postulate models with some (main or interaction) effects ofthe subject variables and evaluate the fit of these models to the data.

    To arrive at the most adequate model, which is a trade-off between goodness-of-fit and parsimony, we followed the procedure of 'forward selection' ofGoodman (Fienberg 1977). This procedure implies that if the minimal modeldoes not fit, first all main effects should be added. If this extended model stilldoes notfit,all possible two-way interactions should be added, and then allthree-way interactions, etc., until the models cannot be improved any further.Once an adequate model (with all k-way interactions) has been achieved, thenone should try to simplify it by deleting one k-way interaction at a time for aslong as it does not affect the fit of the model. This procedure of forward selectionis discussed in the next section ('Results').Some cells in our 2 X 2 X 2 X 6 X 6 table contain no observations, becausethose observations are unlikely (e.g. superord inate of 'secret') and some are justsampling zeros. These cells are left out of the analyses, explaining the deviantnumber of degrees of freedom.RESULTSFirstly, we will have to select a model as the most adequate description of ourfrequency data. Secondly, we will discuss the main and interaction effects whichare postulated in the selected model.Mo del selectionAs Table 1 shows, the minimal model (1) has to be rejected (G2 = 665.91 df =229 p < .05). We must conclude that the frequency data are somewhat morecomplex than this minimal model suggests. The subject variables must have aneffect on the num ber or type of aspects children express.Model 2 postulates main effects for Age, Proficiency, and Ethnic Back-ground, i.e. the different age, proficiency, and ethnic groups are expected todiffer in the number of meaning aspects they express, but these differences areindependent of other variables (no interactions). This second model, though alarge improvement on the minimal model (Gdiff = 328.8 dfdiff = 3 p < .05), isstill not adequate and has to be rejected (G2 = 337.15 df = 226 p < .05). Theconsiderable improvement indicates that the groups must differ strongly as tothe number of meaning aspects expressed. The rejection of the second modelindicates that interactions between variables are necessary in the description ofthe frequency data.

    Model 3 allows for all possible two-way interactions (i.e. interactions betweentwo variables), including the (interesting) interactions between the three subjectvariables and Category. Model 3 is a considerable improvement on the secondmodel. The improvement is statistically significant (Gdiff = 142.5 dfdiff = 33p < .05). The third model in itself cannot be rejected as a description of the data(G2 = 194.62 df= 193 p > .20). Although model 3 could be accepted, wechecked, for certainty, whether model 3 could be improved by adding all

    atOUPsiteacces

    sonJuly31,2012

    http://applij.oxfordjournals.org/

    Downloadedfrom

    http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/
  • 8/22/2019 Lexical Knowledge of monolingual and bilingual children

    12/20

    Table I: Model fit and selection2>po>zzmrm>2D50O03c/>OIoozmz

    Models G7df difference withprevious model1 C + W +C*W2 C + W + E + P + A+C*W34

    C*W*E+ C 'W'P + C*W*A+ C*E*P + C'E* A + C*P*A+ W*E*P + W*E*A+ W*P*A + E*P* A5 C + W + E + P + A+C *W +C* E+ +CA +E*P + E*A

    6 6 5 . 9 l / 2 2 9 t3 3 7 . 1 5 / 2 2 6 t1 9 4 . 6 2 / 1 9 38 0 . 7 8 / 9 3

    2 2 4 . 5 7 / 2 1 4

    n.a.3 2 8 . 8 / 3 t1 4 2 . 5 / 3 3 11 1 3 . 8 / 1 0 0

    n.a.

    C = Category; W = Word; A = Age; P = Proficiency level; E = Ethnic background; G 2 = scaled deviance (approximately chi-square distributed)t statistically significant (p < .05)

    atOUPsiteaccessonJuly31,2012

    http://applij.oxfordjournals.org/

    m

    http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/
  • 8/22/2019 Lexical Knowledge of monolingual and bilingual children

    13/20

    356 LEXICAL KNOW LEDGE IN CHILDRENpossible three-way interactions (model 4). Since model 4 is a more liberal modelthan model 3, it obviously fits the data very well (G2 = 80.78 df = 93 p > .20),but the improvement in fit compared to model 3 is not statistically significant(Gjjff = 113.8 dfdiff = 100 p > .20). So, a model with solely two-way interactionsis an adequate description of the data (model 3) which can hardly be improved.The next question is whether all two-way interactions which are assumed bymodel 3 are relevant, or just some of them. Further model comparisons byleaving out different two-way interactions successively demonstrated that someof the two-way interactions indeed can be left out of the model. The trade-offbetween goodness-of-fit and parsimony is represented by model 5 (see Table 1).EffectsSince we have selected model 5 as the optimal description of our frequencydata, we can now take a closer look at the relevant parameters of the model.Model 5 also includes, besides the (above-mentioned) minimal model para-meters for Word and Category, main effects of the variables Age, Proficiency,and Ethnic Background, the interactions between the subject variables Age andEthnic Background and Category (C*A, C*E), and interactions between EthnicBackground and Proficiency (E*P), and Ethnic Background and Age (E*A). Inthe following, we will discuss the characteristics of these effects. First, we willillustrate the main effects and second the interactions.Main effects. Model 5 assumes main effects of Age, Proficiency, and EthnicBackground. Since these main effects are not Category-related they onlyconcern the number of meaning aspects mentioned. These three main effectsare illustrated in Figures 4 to 6. Figure 4 shows that out of all 8,833 registeredmeaning aspects 58.6 per cent (5,176) come from the Dutch children and 41.4per cent (3,657) from the Turkish children. The effect of Proficiency is smaller.The children considered high proficient (within their group) expressed 52.5 percent (4,637), whereas children considered low proficient expressed 47.5 percent (4,196) of the aspects (Figure 5). The size of the third main effect, that ofAge, is in between the other two: 53.85 per cent (4,757) of the aspects areexpressed by the older children and 46.15 per cent (4,076) by the younger ones(Figure 6). These three main effects are partly mediated by interactions withother variables (see model 5). To get a full picture of the lexical knowledge of thechildren the interaction effects with C ategory will be shown in the following sec-tion.Interaction effects. We were especially interested in the question of whetherdifferent groups (age, proficiency, and ethnic background) use different categor-ies of meaning aspects in their description of word meaning. As m odel 5 indi-cates, Age and Ethnic Background both interact with Category (C*A and C*E).This means that different age groups and different ethnic groups not only pro-duce different numbers of meaning relations, they also produce different typesof relations in defining the words.

    atOUPsiteaccessonJuly31,2012

    http://applij.oxfordjournals.org/

    Downloadedfrom

    http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/
  • 8/22/2019 Lexical Knowledge of monolingual and bilingual children

    14/20

    M A R I A N N E V E R H A L L E N AND ROB SCHOONEN 357

    100 p90 -80 -70 -60 -50 -40 -30 -20 -10 -0

    TurkishDutch

    58.6

    41.4

    Number of aspects mentionedFigure 4: Main effect of Ethnic Background

    100 p90 -80 -70 -60 -50 -40 -30 -20 -10 -0

    I I lowerW> higher

    47.5 52.5

    Number of aspects mentionedFigure 5: Main effect of Proficiency level

    Figure 7 gives the d istribution of meaning aspects over the categories for boththe younger and the older children. The bars in that diagram represent (again)the percentages of the aspects belonging to a certain category. From the maineffect for Age, we already know that the age-groups differ as to the number ofaspects they mention. To make a fair comparison, we equated both the agegroups at 100 per cent for each g roup. Given the main effects, we wanted to seehow often the different types of meaning relations (categories) areused by theyounger and o lder children in their descriptions of word meanings.In Figure 7, we can see the differences between the younger and olderchildren with respect to the kinds of meaning aspects they used. Although bothgroups use syntagmatic relations (Categories 4 and 5; see the section above on'Scoring'for the Category names) most frequently to clarify the meaning of thewords, it shows that the older children tend to use the paradigmatic categories

    atOUPsiteacces

    sonJuly31,2012

    http://applij.oxfordjournals.org/

    Downloadedfrom

    http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/
  • 8/22/2019 Lexical Knowledge of monolingual and bilingual children

    15/20

    358 L E X IC A L K N O W L E D G E IN C H IL D R E N

    lOOp9 0 -8 0 -70 -6 0 -5 0 -40 -30 -20 -10 -0

    young (9 years)old (11 years)

    53.85

    Number of aspects mentionedFigure 6: Main effect of Age

    45403530252015105

    | young (9 years) (100%)| old (11 years) (100%) 3 5 - 8 9 34.9

    26.5222.18

    18.52

    2.18 2.5

    13.89 14.99

    23.4 I 2.99 2.04.OiCat. 1 Cat. 2 Cat. 3 Cat. 4 Cat. 5 Cat. 6Figure 7: Interaction effect of Age by Category

    (Categories 1 to 3) relatively more often than the younger children. The youngerchildren stick relatively more often to the syntagmatic and subjective descrip-tions (Categories 4 to 6).The second interesting interaction effect is between Ethnic Background andCategory. Figure 8 displays a similar pattern to Figure 7. Again, both groupsshow a frequent use of syntagmatic relations in their descriptions. The Dutchchildren, however, use paradigmatic m eaning aspects, especially the importanttaxonomic relations, relatively more often, compared to the Turkish children,who tend to use the syntagmatic and subjective meaning aspects relatively m oreoften. It seems that the differences between the ethnic groups in the types ofmeaning relations expressed parallel the age differences.Besides these two interactions of the Subject variables and C ategory there aretwo more interactions in model 5, one between Ethnic Background and

    atOUPsiteaccessonJuly31,2012

    http://applij.oxfo

    rdjournals.org/

    Do

    wnloadedfrom

    http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/
  • 8/22/2019 Lexical Knowledge of monolingual and bilingual children

    16/20

    MARIANNE VERHALLEN AND ROB SCHOONEN 359

    50%45

    40353025201510

    50

    H I Turkish (100%)S S Dutch (100%)

    22.7617.28

    - 1.97 2.63

    36.12

    27.56

    14.3614.5722.91

    34.81

    2.71 2.32Cat. 1 Cat. 2 Cat. 3 Cat. 4 Cat. 5 Cat. 6

    Figure 8: Interaction effect Ethnic Background by CategoryProficiency (E*P) and one between E thnic Background and Age (E*A). Theseinteractions are not related to the Category variable and thus only refer to thenumber of meaning aspects the children expressed . The differences between theTurkish and Dutch children in the number of meaning aspects expressedinteract with Age and Proficiency level. The difference between the two agegroups is larger for the Turkish group than for the Dutch group. This also goesfor the difference between the proficiency groups, which again is larger for theTurkish group than for the Dutch group.In our discussion, how ever, we will focus on the interactions with Category.DISCUSSIONOur research project has been inspired by the phenomenon that in Dutchprimary education bilingual children appear to profit far less from educationthan monolingual Dutch children. The differences in educational achievementare often attributed to differences in lexical knowledge. Earlier studies haveestablished that during successive school years, bilingual children lag seriouslybehind with regard to the num ber of words they acquire in the second language.This is just one aspect of the problem, however.So far, data have been collectedon the basis of formal vocabulary tests only. To gain a more detailed picture ofthe underlying semantic knowledge of individual words, we used extended worddefinition tasks. The importance of this research method is emphasized bySnow, Cancino, De Temple, and Schley (1991: 90):

    Definitions ... are of both theoretical and practical interest to students of languagedevelopment. In school settings, definitions are often requested of children and givingdefinitions . . . is a standard and frequent technique for vocabulary training. Theoreti-cally, definitions are interesting because they constitute one example of what has beenreferred to as decontextualized language uselanguage used in ways that eschewreliance on sha red social and physical contex t in favour of reliance on a con text c reate dthrough the language itself.

    atOUPsiteacces

    sonJuly31,2012

    http://applij.oxfordjournals.org/

    Downloadedfrom

    http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/
  • 8/22/2019 Lexical Knowledge of monolingual and bilingual children

    17/20

    360 LEXICAL KNOWLEDGE IN CHILDRENWord definitions belong to the domain of what one might call 'intentionallanguage use', demonstrating aspects of language that have a direct bearing oncognition (Bialystok 1991: 8). Formal definitions, however, are a ratherspecialized speech genre, as Snow et al. acknow ledge. Giving a good definitionrequires analysing one's knowledge of word meaning to distinguish between'definitional' and 'incidental information', and also use of the conventional formof giving definitions. The task is still too limited to yield a satisfactory picture of achild's full word knowledge. This is why we extended the definition task: byasking a great num ber of questions, with which m ore meaning d imensions wereexplored. By this procedure, we elicited the expressible knowledge of wordsfrom the children in our project.The results point in the direction of important differences in expressiblelexical knowledge. In the example shown in Table 2, we see the differencesreflected in the responses of a Turkish and a Dutch child to some of the standardquestions on the stimulus word 'nose' (each meaning aspect has been marked bya dash).The differences between the Turkish and Dutch children illustrated in Table2 were similar for all six words. For each word, Turkish children producedfewer meaning aspects and the types of meaning aspects expressed are differentfrom those expressed by the Dutch children, a possible indication of differencesin the underlying lexical semantic system. But ifwe look at children of the Dutchand Turkish subgroups with comparable vocabulary sizes (as measured by theformal test) no differences emerge. Comparable vocabulary size, however, isonly to be found for the nine-year old Dutch low proficient children (belowmedian of their group) and the eleven-year old Turkish high proficient children(above median of their group). An additional loglinear analysis of the data ofthese distant subgroups showed no difference in the number of meaning aspectsexpressed, nor in the types. For these groups the frequency data could bedescribed by the minimal model (C*W). This implies that the most proficienteleven-year old Turkish children eventually reach a level in expressingmeanings comparable with that of low-achieving Dutch children at an earlierage.Of course, our findings only concern the L2 component of the lexicon of thebilingual children. To get a full picture of their underlying lexical semanticsystem we also have to describe the responses of the children in the Turkishinterviews (Verhallen and Schoonen in preparation). Nevertheless, whenconsidered from the perspective of educational achievement and educationalopportunities, the level of achievement in Dutch is of the highest importance. A tschool, Dutch is used exclusively in the regular teaching/learning situations . LIknowledge is of minor im portance for profiting from the lessons. So it is relevantto consider the Dutch lexical knowledge of bilingual children in its own right.What, then, are the consequences of our findings for formal education? Ifbilingual children have a less-developed command of various meaning aspectsof words, teachers cannot ignore these lexical problem s. In the context of formaleducation, the development of (hierarchical) paradigmatic m eaning assignment

    atOUPsiteacces

    sonJuly31,2012

    http://applij.oxfordjournals.org/

    Downloadedfrom

    http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/
  • 8/22/2019 Lexical Knowledge of monolingual and bilingual children

    18/20

    MARIANNE VERHALLEN AND ROB SCHOONEN 361

    Table 2Quest ions

    Wat is een neus?(What is a nose? |

    Hoe zou je uitleggen wat eenneus is?|How would you explain whata nose is? |

    Wat zie je aan een neus?|what do you see if you look ata n o s e ? |Wat voor soorten zijn er?|what kinds of noses arethe re? |

    Wat voor soort ding is eenneus?[what kind of thing is a no se ?|

    Wat kun je doen met je neus?|what can you do with a nose?|

    Turkish child9-years old- met je neus kan je ruiken

    [you can smell with yo urnose |

    |repeats the answer above|

    - het is puntig|it is pointed|

    - grote neus[big nose |

    - kleine neus|small nose|

    - het is een ne us|it is a nose|

    -je kunt het wassen[you can wash it]

    - je neus snuiten|blow your nose|

    Dutch child9-years old- een neus is een lichaamsdeel

    |a nose is a part of the b ody |- je kunt ermee ruiken

    [you can smell with it|- en ademen

    |and breathe|- het heeft een botje|it has a bon e|- het heeft twee gaatjes

    |it has (wo little ho les |- en een velletje

    |and skin |- het is puntig

    |it is pointed|

    - wipneus|snub nose|

    - haakneus|hook nose|

    - dieren neusen[animal noses|

    - het is een ruikdin g|it is a thing of smell|

    - je kunt hem snuiten[you must blow your nose|

    - en je neus afvegen|and wipe your nose|

    - of snuiven|or sniff]

    is of special im portance, because it can be seen as directly linked to educationaldevelopment. The differences we found in the two age groups are consistentwith this claim: older children use paradigmatic categories relatively more oftenthan younger children. The importance of paradigmatic meanings and thedevelopment of a hierarchical system is emphasized by Vygotsky (1962: 92)who states that:

    atOUPsiteaccessonJuly31,2012

    http://applij.oxfo

    rdjournals.org/

    Do

    wnloadedfrom

    http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/
  • 8/22/2019 Lexical Knowledge of monolingual and bilingual children

    19/20

    362 LEXICAL KNOWLEDGE IN CHILDRENa concept can become subject to consciousness and deliberate control only when it is apart of a system. If consciousness means generalization, generalization in turn meansthe forma tion of a sup ero rdin ate conc ept that includes the given con cept as a partic ularcase. A supero rdina te conce pt implies the existence of a series of subordin ate c once pts,and it also presupposes a hierarchy of concepts of different levels of generality. Thusthe given co nce pt is placed within a system of relations hips of generality.

    In the meaning aspects expressed by the bilingual children in our study, theseparadigmatic relations are underrepresented . It is important for teachers to payattention to the fact that lexical delay in bilingual children is not confined to thenumber of L2 words. Looking at words like 'nose' that seem to have beenacquired completely, we must conclude that there a re important restrictions inthe number and in the range of meaning aspects that bilingual children canexpress. The fact that a child produces a word does not mean that he or she usesthe word in all its conceptual implications. Teachers must be continuouslyaware that there are missing links, gaps, or fuzzy relations in immigrantchildren's lexical systems that may not be immediately noticeable, but cannottherefore be ignored.(Revised version received December 1992)

    ACKNOWLEDGEMENTSThe authors want to thank Engelien Verhoeff, Rene A ppel, Rob G erritsen, andFolkert Kuiken for their careful reading of an earlier draft and their helpfulcomments.NOTES

    1 Due to ties, the subdivision according to proficiency did not result in two groups ofequal size, but in a group of 9 and a group of 11 children. However, data werestandardized to group size 10.2 For stimulus words with missing data (like roofdier and wekker), the samestandardization proc edu re (see note 1) was followed.3 The exact statistical distribution of G 2 is unknown; but in large samples, G 2 isapproximately chi-square distributed, so we will use the chi-square tables for statisticaltesting.

    REFERENCESAnglin, J. M . 1985 . 'The child's expressible know ledge of wo rdco nce pts ' in K. E. Nelson

    (ed.) 1985: Children's Language: Volume 5. New Jersey: Lawrence Erlbaum.Barclay, J. R., J. D. Bransford, J. J. Franks, N. S. McCarrell, and K. Nitsch. 1974."Comprehension and semantic flexibility.' Journal of Verbal Learning and Verbal

    Behaviour 13:47 '1-8 1 .Bialystok, E. 19 91 . "Introduction ' in E. Bialystok (ed.) 1 99 1: Language Processing inBilingual Children. Cambridge: Cambridge University Press.

    Beck, I. L., M. G. McKeown,and R. C. Omanson. 19 87 .'T he effects and uses of diversevocabulary instructional technique s' in M. G. McK eown and M. E. Curtis (eds.) 1987:Fhe Nature of Vocab ulary Acquisition. New Jersey: Lawrence Erlbaum .

    atOUPsiteacces

    sonJuly31,2012

    http://applij.oxfordjournals.org/

    Downloadedfrom

    http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/
  • 8/22/2019 Lexical Knowledge of monolingual and bilingual children

    20/20

    MARIANNE VERHALLEN AND ROB SCHOONEN 363Boogaard, M., R. Damhuis, K. De Glopper, and H. Van den Berg. 1990. De Neder-

    landse Taa lvaardigheid van Allochtone en Nederlandse kleuters [Dutch LanguageProficiency of Immigrant and Dutch Children). Forum 4. Lisse: Swets and Zeitlinger.Cronbach, L. J. 194 2. 'An analysis of techniq ues for diagno stic vocabulary testing.' Jour-nal of Educational Research 3 6 : 2 0 6 - 1 7 .Cruse, D. A. 1986. Lexical Semantics. Cambridge: Cambridge University Press.Dickinson, D. 1984. 'First impressions: Children's knowledge of words gained from a

    single expo sure. ' Applied Psycholinguistics 5: 3 5 9 - 7 3 .Everitt, B. S. 1976. The Analysis of Contingency Tables. London: Chapman and Hall .Fienberg, S . E. 1 9 77 . The Analysis of Cross-Classified Categorical Data. Cambr idge ,

    MA : MIT Press.Kuczaj, S . A. 1982. 'Acquisition of word meaning in the context of the development ofthe semantic system' in C. J. Brainerd and M. Pressley (eds.) 1982: Verbal Processes in

    Children. New York: Springer Verlag.Miller, G. A. and P. N. Johnson-Laird. 1976 . Language and Perception. Cambr idge ,

    MA: Harvard University Press.Nagy, W. E., P. A. Herman, and R. C. Anderson. 198 5. 'Learning w ords from context. '

    Reading Research Quarterly 2 0 / 2 : 2 3 3 - 5 3 .Snow, C . E., H. Cancino, J. D e Tem ple, and S. Schley. 1991 . 'Giving formal d efinitions'in E. Bialystok (ed.) 1991: Language Processing in Bilingual Children. Cambridge:Cambridge University Press.

    Sternberg, R. J. 198 7. 'Most vocab ulary is learned from context ' in M. G. M cKeow n andM. E. Curtis (eds.) 1987 : The Nature of Vocabulary Acquisition. New Jersey:Lawrence Erlbaum.

    Verhoeven, L. and A. Vermeer. 1989 . Diagnose van Kindertaal [Diagnosis of ChildLanguage]. Tilburg: Zwijssen.Vygotsky, L. 1962. Thought and Language. Cam bridge, M A: M IT Press.Watson, R. 1985. 'Towards a theory of definition.' Journal of Child Language 12: 181

    97.atOUPsiteacces

    sonJuly31,2012

    http://applij.oxfo

    rdjournals.org/

    Do

    wnloadedfrom

    http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/http://applij.oxfordjournals.org/